Skip to content

Commit

Permalink
Merge pull request #7 from Open-Speech-EkStep/v2-hydra
Browse files Browse the repository at this point in the history 
Release Torchscript Models
  • Loading branch information
@harveenchadha
harveenchadha authored Sep 4, 2021
2 parents 4b9b010 + e8aad97 commit 78f5667
Showing 1 changed file with 14 additions and 12 deletions.
26 changes: 14 additions & 12 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,17 +14,16 @@

## Finetuned Models (works on [v2-hydra](https://github.com/Open-Speech-EkStep/vakyansh-wav2vec2-experimentation/tree/v2-hydra) branch*)

| Language | Pretrained Model | Finetuned Model | Dictionary | Single Model for Inference | Finetuned Hours | wandb logs |
|----|--------|----|-----|-------------------|-------|----|
| Hindi | CLSRIL-23 | [him_4200](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/him_4200.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/dict.ltr.txt ) | [hindi_infer](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/hindi_infer.pt) | 4200 h | [hindi_logs](https://wandb.ai/harveenchadha/hindi_finetuning_multilingual) |
| Indian English | CLSRIL-23 | [enm_700](https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/enm_700.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/dict.ltr.txt) | [english_infer]( https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/english_infer.pt ) | 700 h | [english_logs](https://wandb.ai/harveenchadha/finetuning_english_multilingual) |
| Kannada | CLSRIL-23 | [knm_560](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/knm_560.pt) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/dict.ltr.txt) | [kannada_infer](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/kannada_infer.pt ) | 560 h | [kannada_logs](https://wandb.ai/priyanshi_shah/kannada_finetuning_multilingual) |
| Tamil | CLSRIL-23 | [tam_250](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/tam_250.pt) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/dict.ltr.txt) | [tamil_infer](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/tamil_infer.pt ) | 250 h | [tamil_logs](https://wandb.ai/harveenchadha/tamil-finetuning-multilingual) |
| Bengali | CLSRIL-23 | [bnm_200](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/bnm_200.pt) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/dict.ltr.txt) | [bengali_infer](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/bengali_infer.pt ) | 200 h | [bengali_logs](https://wandb.ai/harveenchadha/bengali-finetuning-multilingual) |
| Nepali | CLSRIL-23 | [nem_130](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/nem_130.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/dict.ltr.txt) | [nepali_infer]( https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/nepali_infer.pt ) | 130 h | [nepali_logs](https://wandb.ai/harveenchadha/nepali-finetuning-multilingual) |
| Telugu | CLSRIL-23 | [tem_100](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/tem_100.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/dict.ltr.txt ) | [telugu_infer](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/telugu_infer.pt ) | 100 h | [telugu_logs](https://wandb.ai/harveenchadha/telugu-finetuning-multilingual) |
| Gujarati | CLSRIL-23 | [gum_100](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/gum_100.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/dict.ltr.txt ) | [gujarati_infer](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/gujarati_infer.pt ) | 100 h | [gujarati_logs](https://wandb.ai/harveenchadha/gujarati-finetuning-multilingual) |

| Language | Pretrained Model | Finetuned Model | Dictionary | Single Model for Inference | Finetuned Hours | wandb logs | TS model |
|----|--------|----|-----|-------------------|-------|----|---|
| Hindi | CLSRIL-23 | [him_4200](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/him_4200.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/dict.ltr.txt ) | [hindi_infer](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/hindi_infer.pt) | 4200 h | [hindi_logs](https://wandb.ai/harveenchadha/hindi_finetuning_multilingual) | [hindi_ts](https://storage.googleapis.com/vakyansh-open-models/models/hindi/hi-IN/vakyansh-wav2vec2-hindi-him-4200_quant.pt)
| Indian English | CLSRIL-23 | [enm_700](https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/enm_700.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/dict.ltr.txt) | [english_infer]( https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/english_infer.pt ) | 700 h | [english_logs](https://wandb.ai/harveenchadha/finetuning_english_multilingual) | [english_ts](https://storage.googleapis.com/vakyansh-open-models/models/english/en-IN/vakyansh-wav2vec2-indian-english-enm-700_quant.pt)
| Kannada | CLSRIL-23 | [knm_560](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/knm_560.pt) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/dict.ltr.txt) | [kannada_infer](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/kannada_infer.pt ) | 560 h | [kannada_logs](https://wandb.ai/priyanshi_shah/kannada_finetuning_multilingual) | [kannada_ts](https://storage.googleapis.com/vakyansh-open-models/models/kannada/kn-IN/vakyansh-wav2vec2-kannada-knm-560_quant.pt)
| Tamil | CLSRIL-23 | [tam_250](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/tam_250.pt) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/dict.ltr.txt) | [tamil_infer](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/tamil_infer.pt ) | 250 h | [tamil_logs](https://wandb.ai/harveenchadha/tamil-finetuning-multilingual) | [tamil_ts](https://storage.googleapis.com/vakyansh-open-models/models/tamil/ta-IN/vakyansh-wav2vec2-tamil-tam-250_quant.pt)
| Bengali | CLSRIL-23 | [bnm_200](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/bnm_200.pt) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/dict.ltr.txt) | [bengali_infer](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/bengali_infer.pt ) | 200 h | [bengali_logs](https://wandb.ai/harveenchadha/bengali-finetuning-multilingual) | [bengali_ts](https://storage.googleapis.com/vakyansh-open-models/models/bengali/bn-IN/vakyansh-wav2vec2-bengali-bnm-200_quant.pt) |
| Nepali | CLSRIL-23 | [nem_130](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/nem_130.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/dict.ltr.txt) | [nepali_infer]( https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/nepali_infer.pt ) | 130 h | [nepali_logs](https://wandb.ai/harveenchadha/nepali-finetuning-multilingual) | [nepali_ts](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/vakyansh-wav2vec2-nepali-nem-130_quant.pt)
| Telugu | CLSRIL-23 | [tem_100](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/tem_100.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/dict.ltr.txt ) | [telugu_infer](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/telugu_infer.pt ) | 100 h | [telugu_logs](https://wandb.ai/harveenchadha/telugu-finetuning-multilingual) | [telugu_ts](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/vakyansh-wav2vec2-telugu-tem-100_quant.pt)
| Gujarati | CLSRIL-23 | [gum_100](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/gum_100.pt ) | [dict](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/dict.ltr.txt ) | [gujarati_infer](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/gujarati_infer.pt ) | 100 h | [gujarati_logs](https://wandb.ai/harveenchadha/gujarati-finetuning-multilingual) | [gujarati_ts](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/vakyansh-wav2vec2-gujarati-gnm-100_quant.pt)

*\*Note: All future models will be open sourced on v2-hydra branch*

Expand All @@ -44,7 +43,10 @@ Data is taken from [AI For Bharat Corpus](https://indicnlp.ai4bharat.org/corpora
| Nepali | kenlm 5-gram | [nepali_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/lexicon.lst ) | [nepali_lm](https://storage.googleapis.com/vakyansh-open-models/models/nepali/ne-IN/lm.binary ) |
| Telugu | kenlm 5-gram | [telugu_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/lexicon.lst) | [telugu_lm](https://storage.googleapis.com/vakyansh-open-models/models/telugu/te-IN/lm.binary) |
| Gujarati | kenlm 5-gram | [gujarati_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/lexicon.lst ) | [gujarati_lm](https://storage.googleapis.com/vakyansh-open-models/models/gujarati/gu-IN/lm.binary) |

| Marathi | kenlm 5-gram | [marathi_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/marathi/mr-IN/lexicon.lst) | [marathi_lm](https://storage.googleapis.com/vakyansh-open-models/models/marathi/mr-IN/lm.binary) |
| Odia | kenlm 5-gram | [odia_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/odia/or-IN/lexicon.lst) | [odia_lm](https://storage.googleapis.com/vakyansh-open-models/models/odia/or-IN/lm.binary) |
| Sanskrit | kenlm 5-gram | [sanskrit_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/sanskrit/sa-IN/lexicon.lst) | [sanskrit_lm](https://storage.googleapis.com/vakyansh-open-models/models/sanskrit/sa-IN/lm.binary)
| Maithili | kenlm 5-gram | [maithili_lexicon](https://storage.googleapis.com/vakyansh-open-models/models/maithili/mai-IN/lexicon.lst) | [maithili_lm](https://storage.googleapis.com/vakyansh-open-models/models/maithili/mai-IN/lm.binary)


## Domain Specific Language Models
Expand Down

0 comments on commit 78f5667

Please sign in to comment.